Impact of Recreational Cannabis Legalization on Opioid Prescribing and Opioid-Related Hospital Visits in Colorado: an Observational Study.

Background: Cannabis may be a substitute for opioids but previous studies have found conflicting results when using data from more recent years. Most studies have examined the relationship using state-level data, missing important sub-state variation in cannabis access.

Objective: To examine cannabis legalization on opioid use at the county level, using Colorado as a case study. Colorado allowed recreational cannabis stores in January 2014. Local communities could decide whether to allow dispensaries, creating variation in the level of exposure to cannabis outlets.

Design: Observational, quasi-experimental design exploiting county-level variation in allowance of recreational dispensaries.

Subjects: Colorado residents MEASURES: We use licensing information from the Colorado Department of Revenue to measure county-level exposure to cannabis outlets. We use the state's Prescription Drug Monitoring Program (2013-2018) to construct opioid-prescribing measures of number of 30-day fills and total morphine equivalents, both per county resident per quarter. We construct outcomes of opioid-related inpatient visits (2011-2018) and emergency department visits (2013-2018) with Colorado Hospital Association data. We use linear models in a differences-in-differences framework that accounts for the varying exposure to medical and recreational cannabis over time. There are 2048 county-quarter observations used in the analysis.

Results: We find mixed evidence of cannabis exposure on opioid-related outcomes at the county level. We find increasing exposure to recreational cannabis is associated with a statistically significant decrease in number of 30-day fills (coefficient: -117.6, p-value<0.01) and inpatient visits (coefficient: -0.8, p-value: 0.03), but not total MME nor ED visits. Counties with no medical exposure prior to recreational legalization experience greater reductions in the number of 30-day fills and MME than counties with prior medical exposure (p=0.02 for both).

Conclusions: Our mixed findings suggest that further increases in cannabis beyond medical access may not always reduce opioid prescribing or opioid-related hospital visits at a population level.